It seems New York has decided it’s not waiting for Washington to get its act together on artificial intelligence. While federal regulators are still debating the rulebook, Governor Kathy Hochul has just signed the RAISE Act into law, catapulting New York to the forefront of AI safety regulation in the United States, right alongside California. This isn’t just another bit of legislative paperwork; it’s a clear signal that the era of AI development operating in a regulatory free-for-all is well and truly over. So, what exactly has been signed, and why are some of the biggest names in tech both applauding and quietly fighting it?
So, What Is This RAISE Act Anyway?
At its core, the RAISE Act (which stands for Reporting on AI Safety and Effectiveness) is designed to force a new level of transparency and accountability on the developers of the most powerful AI models. Think of it like food safety regulations. We don’t just trust a food manufacturer’s word that their product is safe; we require them to list ingredients and follow specific hygiene protocols. The RAISE Act applies a similar logic to large-scale AI.
The main objective is to prevent disasters before they happen by making sure developers are not only building safety checks into their systems but are also willing to be open about them. Two key pillars uphold this framework: safety protocol disclosure and timely AI incident reporting. For years, the inner workings of models from companies like OpenAI and Google have been a black box. This legislation aims to install a window into that box, and as Governor Hochul put it, create a “unified benchmark among the country’s leading tech states” while the federal government lags.
The Fine Print: Unpacking the RAISE Act Requirements
The devil is always in the details, and the requirements of this act are pointedly specific. It’s not about vague promises of “ethical AI”; it’s about concrete, mandated actions that have real teeth.
Show Me Your Homework: Mandatory Disclosure of Safety Protocols
First and foremost, the act mandates that developers of powerful “covered” models must disclose their safety testing protocols. This isn’t a suggestion; it is a legal requirement. These developers will have to publicly share the results of the safety tests they conduct on their models before they are released.
What does this mean in practice? It forces a culture of responsible AI development by making safety a public-facing metric, not just an internal R&D memo. Companies can no longer simply claim their AI is safe; they have to prove it with data. This disclosure is designed to show how they tested for risks like bias, misinformation generation, and potential for malicious use. It’s a fundamental shift from a “move fast and break things” mentality to a “prove it’s safe before you ship it” approach.
The 72-Hour Fire Drill: Timely Incident Reporting
The second major component is perhaps the most operationally challenging for AI labs: the requirement to report significant safety incidents to the state within 72 hours. According to the TechCrunch article that broke down the bill, this is one of the strictest reporting timelines in any tech regulation to date.
An “incident” could be anything from a model consistently producing dangerous information to being exploited for a large-scale cyber-attack. This rapid reporting timeline ensures that regulators aren’t finding out about major issues from the press weeks later. It allows them to react quickly, warn the public if necessary, and hold the company accountable. It creates a direct line of communication between the AI lab and the state, a crucial link in any real safety framework.
Follow the Money: Enforcement and Penalties
A law without consequences is just a suggestion. New York has made it clear they are serious by assigning enforcement to a new, dedicated office within the Department of Financial Services (DFS) and backing it up with significant financial penalties.
The New Sheriff in Town
Why the DFS? Because it has a long history of regulating complex, high-stakes industries like banking and insurance. It understands risk management and has the investigative muscle to audit powerful corporations. This move signals that New York sees AI not just as a technology but as a critical piece of economic and social infrastructure that requires serious oversight.
New York isn’t alone, but it is moving assertively. California was the first state to pass a similar law, and as Anthropic’s Head of Global Policy, Sarah Heck, noted, “The fact that two of the largest states have now enact AI transparency legislation signals the critical importance of safety.” This state-level action creates a de facto national standard, putting pressure on other states and the federal government to harmonise their rules.
Hitting Them Where It Hurts
The penalties for non-compliance are severe enough to get any board of directors’ attention. The law proposes:
– A fine of up to $1 million for the first violation.
– A fine of up to $3 million for any subsequent violations.
For behemoths like OpenAI and Anthropic, these sums might not be business-ending, but they are far from trivial. More importantly, they create a significant financial and reputational risk. No company wants to be on the receiving end of a multi-million-dollar fine for failing to report a safety breach. It’s a powerful incentive to get their internal processes in order fast.
The Industry’s Double Game
Here’s where it gets interesting. Publicly, many major AI companies have expressed support for the Kathy Hochul regulation. OpenAI and Anthropic, for instance, gave statements supporting the bill’s focus on safety. But behind the scenes, a different story was unfolding.
A super PAC backed by big-name venture capital firm Andreessen Horowitz and OpenAI’s own co-founder, Greg Brockman, was reportedly lobbying heavily to kill or weaken the bill. State Senator Andrew Gounardes, a sponsor of the bill, didn’t mince his words: “Big Tech thought they could weasel their way into killing our bill. We shut them down and passed the strongest AI safety law in the country.” This split between public posture and private lobbying reveals the deep tension within the industry. They want to be seen as champions of safety, but they also want as few operational constraints as possible.
What Happens Next? The Road Ahead
With the bill now signed, the real work begins. Governor Hochul has already agreed with lawmakers to revisit the bill next year to make “changes,” a concession likely made to get the bill over the line despite the intense lobbying. What those changes will look like is anyone’s guess, but they will be the next battleground for proponents of strong regulation and industry lobbyists.
This also sets up a potential clash with federal policy. Former President Donald Trump recently issued an executive order directing federal agencies to challenge state-level AI laws, creating a patchwork of conflicting rules. This push-and-pull between state innovation and federal oversight will define the next chapter of American tech regulation. Will a national standard emerge, or will we see a splintered landscape where developers have to navigate different rules in New York, California, and Texas?
The RAISE Act is a landmark piece of legislation. It codifies a set of powerful ideas about responsible AI development and makes New York a crucial player in the global conversation. While it may be tweaked and refined, the core RAISE Act requirements have established a new baseline for accountability. The days of building world-changing technology in secret are numbered.
What do you think? Is this kind of state-level regulation the right way to ensure AI safety, or does it risk creating a confusing and burdensome system for developers?
